llm 0.37.2 fixes accidental OpenAI API key disclosure
Original title: llm 0.37.2
A key could be sent if api_key_name was never stored, or a responses model ran with chat completions forced.
The llm 0.37.2 release fixes a security issue reported by Daniel de Araujo in GHSA-v9h7-jj4g-f862 and GHSA-r639-rr7r-x72w. LLM can call OpenAI-compatible endpoints from other providers through extra-openai-models.yaml, now managed by the llm openai endpoint add command. De Araujo reported two ways a configured OpenAI API key could accidentally be sent to one of those providers.
Unconfirmed: The report does not say whether any API keys were actually sent to third-party providers.
- Advisories cited are GHSA-v9h7-jj4g-f862 and GHSA-r639-rr7r-x72w.
- One path involved an api_key_name whose key had never been stored.
- The other used a responses: true model with -o chat_completions 1.
- Endpoint configuration is now managed by llm openai endpoint add.
Impact: Configured OpenAI API key sent to another provider
| Product | Affected | Fixed in |
|---|---|---|
| Simon Willison llm | — | 0.37.2 |
- 1.Upgrade to llm 0.37.2.
- Products
- llm
From the original article
Release: llm 0. 37. 2 Fixes a security issue reported by Daniel de Araujo in GHSA-v9h7-jj4g-f862 and GHSA-r639-rr7r-x72w .
simonwillison.net
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.